jleo, v31 n3 firms’ qualifications and subcontracting in

Firms’ Qualifications and Subcontracting in Public

Procurement: An Empirical Investigation

Luigi Moretti*

University of Padova

Paola Valbonesi**

University of Padova & NRU-HSE

Using a newly assembled dataset, we empirically investigate the effects of

subcontracting on procurement auction prices in Italy. In this setting, the pre-

qualifications required for firms aiming to bid on public contracts determine the

firms’ different subcontracting formats. We find that fully qualified firms in a

position to choose whether to subcontract generally offer lower prices than par-

tially qualified firms, which must proceed with mandatory subcontracts. This

result indicates that the firms’ voluntary arrangements tend to improve market

performance, while imposed arrangements tend to worsen market performance,

in the public procurement supply-chain. (JEL H57, L23, L24, D44)

*Senior Research Fellow at the Department of Economics and Management, University

of Padova. Email: [email protected].

**Associate Professor of Economics at the Department of Economics and Management,

University of Padova and Research Professor at the National Research University Higher

School of Economics (NRU-HSE). Email: [email protected].

We are particularly indebted to StephenMartin for his helpful suggestions. We are grate-

ful to the Editor and two anonymous referees for their valuable comments on the previous

version of the paper. We also wish to thank Klenio Barbosa, Alessandro Bucciol, Andrea

Camanzi, Decio Coviello, Ottorino Chillemi, Francesco Decarolis, Gianni De Fraja,

Dakshina G. De Silva, Caterina Giannetti, Ricard Gil, Stefano Galavotti, Roberto

Grandinetti, Elisabetta Iossa, Francine Laffontaine, Marco Pagnozzi, Giancarlo Spagnolo,

Yossi Spiegel, Steve Tadelis, Piero Tedeschi, and Francesca Zantomio, as well as the partici-

pants at the International Conference on “Public procurement and sustainable growth”, San

Servolo (Venice, 20–21 October 2011); the International Conference “Contracts, procure-

ment, and public-private arrangements”, Paris (30–21 May 2012); the Workshop PRIN on

“The new industrial policy and private-public interactions between firms andmarkets”, Capri

(5–6 July 2012); the XXXIX EARIE Conference, Rome (2–4 September 2012); the XXVII

Jornadas of economia industrial, Murcia (13–14 September 2012); the Conference of the

Italian Society of Economists (SIE), Matera (18–20 October 2012); the 2013 Conference of

International Industrial Organization Society, Boston (17–19 May 2013); the Workshop on

“Sustainable public procurement: research trends and new challenges”, HSE-NRUMoscow

(4�5 October 2013); and the Department’s seminars in Padova (June 2012) and Pisa (April

2012) for their helpful comments and suggestions.We are also grateful to the Osservatorio sui

Contratti Pubblici, AVCP (Filippo Romano, Alberto Zaino) and the Ufficio Appalti e

Lavori, Regione Valle d’Aosta (Edmund Freppa, Marisa Covolo) for providing data. We

gratefully acknowledge the financial support of the ItalianMinistry of Education, University

and Research (grant PRIN2006-2006130472 003) and the University of Padova (grant N.

CPDA084881/08). All remaining errors are our own.

The Journal of Law, Economics, and Organization, Vol. 31, No. 3doi:10.1093/jleo/ewv001Advance Access published February 11, 2015� The Author 2015. Published by Oxford University Press on behalf of Yale University.All rights reserved. For Permissions, please email: [email protected]

JLEO, V31 N3 568D

ownloaded from

https://academic.oup.com

/jleo/article/31/3/568/808684 by Universita' Padova BC

I user on 21 Decem

ber 2020

1. Introduction

Subcontracting usually involves “a reallocation of production require-

ments among firms” (Kamien and Li 1990: 1354), a process that is partof a firm’s strategic production planning. Many theoretical contributions

have addressed the determinants of subcontracting versus vertical integra-

tion and, more generally, the determinants of the firm’s boundaries based

on transaction costs (Williamson 1985), property rights (Grossman and

Hart 1986), and the knowledge-based view of the firm (Kogut and Zander1992, 1996). Empirical research has provided several case studies that have

documented how and when firms adopt subcontracting to efficiently or-

ganize production in different economic sectors.1

These theoretical and empirical contributions are all based on the firm’s

voluntary choice of internal/external sourcing. In public procurement,however, firms are often constrained by many rules that limit their

decision-making abilities. What decisions should firms be allowed to

make in production planning to efficiently execute a public contract?

The answer to this question should impinge on the policy makers’

design of public procurement procedures. Indeed, rigid rules might limitthe firms’ decisions regarding the supply chain and thus affect the effi-

ciency of procurement transactions and overall social welfare.Typically, the regulatory burden related to the decision to subcontract

in the execution of public contracts can be explained by two main factors.

First, public resources allocated through these procurement contracts areoften specifically intended to be affirmative action policies that indirectly

enhance the participation of disadvantaged business enterprises (DBEs)

through subcontracting schemes. Second, highly regulated procedures are

adopted to maintain fairness in the procurement transactions in order to

prevent favoritism, collusion, corruption, and/or poor performance.Concerning the first factor, empirical evidence of the effect of rules

requiring the participation of DBEs as subcontractors on the procurement

costs is by no means conclusive.2 Marion (2009) exploited a modification

of the Californian law that eliminated a preferential treatment policy and

found that the average price of procured items fell by 5.6% in the state’sDepartment of Transportation contracts. De Silva et al. (2012) empirically

investigated the effect of a subcontracting goal program in Texas and

compared projects in which prime contractors were obliged to outsource

1. A seminal empirical survey on the firms’ vertical arrangements versus spot market

transactions and long term contracts in different sectors has been proposed by Joskow

(1988); more recently, Lafontaine and Slade (2007) have provided a thoughtful empirical

survey on backward and forward vertical integration. Several case studies specifically inves-

tigate outsourcing in different sectors; see, among others, Novak and Stern (2008) and

Macher (2006) on the automobile and semiconductor industries, respectively. Finally, for a

discussion on core competencies and activities that are better performed externally, see Quinn

and Hilmer (1994).

2. DBE programs are also implemented as bid preference schemes or set-aside auctions

for such business (Marion 2007; Krasnokutskaya and Seim 2011; Athey et al. 2013).

Firms’ Qualifications and Subcontracting in Public Procurement 569D

ownloaded from



I user on 21 Decem

ber 2020

a portion of their contracts to DBEs with projects in which they were not

obliged to do so; they found little differences between the projects in the

level of submitted bids.3

Concerning the second factor, relevant examples are provided by frame-

works for supplier qualification screening, which typically aim to verify

that the supplier is indeed able to comply with all of the contract specifi-

cations with a reasonable degree of certainty. In this aim, supplier quali-

fication screening often involves the verification of a firm’s financial status,

references, product, and surge capacity. These procedures determine a

firm’s entry into the public procurement market and, according to the

adopted rules and verification criteria, can directly affect the supplier’s

make-or-subcontract choice when executing a contract. Although supplier

qualification screening is commonly used in many countries, it is surpris-

ing that so little has been written about the effects of such procedures on

the procurement costs.4 To the best of our knowledge, no empirical work

has investigated how the rules regarding firm qualification screening affect

the subcontracting and, hence, the public contracts in terms of both prices

and performed quality.5 This dearth of research is surprising, considering

that public procurement accounts for approximately 15% of the GDP in

developed countries. Thus, given the size of the market, it is necessary to

have a clear understanding of the best rules that foster the contractors’

ability to make efficient supply-chain decisions.6

This article aims to help fill this research gap. We empirically investigate

the Italian regulations concerning qualification screening for public works

that determine two different positions regarding a firm’s decision to sub-

contract or not. Similar to regulations in other national procurement

settings,7 the Italian regulations for public works require that: (1) sup-

pliers undergo a preliminary qualification screening before they enter the

public contract market, and (2) every task in public contracts be

3. As suggested byDe Silva et al. (2012), these contrasting results can be driven by the very

different environments that the two studies focused on.

4. Experimental and theoretical studies comparing costly ex-ante or ex-post qualification

screening in procurement have been recently provided byWan et al. (2012) andWan and Beil

(2009).

5. Conversely, a conspicuous number of empirical contributions have investigated which

auction format should be addressed by procurement regulation to promote lower awarding

prices and better contractual performances. See, among others, Bajari et al. (2009), Decarolis

(2013), Bucciol et al. (2013), Lewis and Bajari (2011), and Olivares et al. (2012).

6. The recent “Green Paper on the modernization of EU public procurement policy”

(2011) indicates that subcontracting is a relevant tool to encourage the participation of

small and medium enterprises (SMEs) in public procurement contracts, whereby SMEs are

considered to be of crucial importance for stimulating job creation, economic growth, and

innovation. However, no recommendations for best practices are provided.

7. Qualifications are needed to enter the market for public contracts in many EU coun-

tries, the United States, and Japan; however, the design of these systems and the criteria

adopted differ somewhat. For a few detailed examples, see OECD (2007).

570 The Journal of Law, Economics, & Organization, V31 N3D

ownloaded from



I user on 21 Decem

ber 2020

completed by qualified suppliers. Specifically, these requirements affect a

firm’s make-or-subcontract decision in two ways:

. if the firm is not qualified to complete all of the tasks in a given contract,

its production strategy has to take into account a “mandatory sub-

contracting” agreement with another firm that is qualified to do so;

. if the firm is fully qualified to complete all of the tasks in a given con-

tract, its production strategy may or may not involve subcontracting

for part of the work to a similarly qualified firm; and, in this case, the

firm holds an “optional subcontracting” position.

Thus, qualified firms bid for each tendered public contract with the

advance knowledge that they are either in a position to opt for subcon-

tracting or required to do so once they win the contract. Our aim is to test

whether these subcontracting positions are likely to affect a firm’s bid and

thus to determine different procurement costs. We also control for the

effects of optional and mandatory subcontracting on ex-post contract

performance; that is, the probability of time and cost overrun.The data. We have assembled an original dataset containing informa-

tion on the Italian public procurement contracts awarded by open tender

and the characteristics of the bidding firms. For each tendered contract,

we specifically collected information regarding the tasks to be completed

(i.e., the “categories of work” corresponding to the qualifications

required), the identity and the qualifications of the bidding firm. Thus,

by matching the qualifications that are required to execute a contract with

each bidder’s qualifications, we are able to identify the bids that are

offered by partially qualified firms (i.e., firms that will have to engage in

mandatory subcontracting if they win the auction) and by fully qualified

firms (i.e., firms that may choose to complete the project by themselves or

to use subcontractors).Our identification and results. We adopted a reduced-form approach

and control for auction/contract-related characteristics and for firm char-

acteristics and fixed effects. In fact, our identification relied on the within-

firm variability of the subcontracting position when bidding for different

contracts. In our sample, most of the bids (74.7%) were offered by firms

that had taken part both in auctions where they had an optional subcon-

tracting position and in auctions where they had a mandatory subcon-

tracting position. This approach allowed us to reduce the omitted

variables problem and to limit differences in the types of firms that

adopted optional or mandatory subcontracting. Furthermore, in our

aim to better capture the firm characteristics that can also vary over

time, in different specifications of the model we controlled for the firm-

year-fixed effects, which allowed us to focus on the within-firm and firm-

year variations in the subcontracting positions.Our empirical results show that bidding firms in a position to choose

whether to subcontract part of the work (i.e., optional subcontracting)


ownloaded from



I user on 21 Decem

ber 2020

offer lower prices than those obliged to subcontract part of the work (i.e.,mandatory subcontracting). This effect is still significant when we focus onthe sub-sample of bids offered by winning firms that actually did engage insubcontracting. These findings indicate that the production efficiencyderived from subcontracting is higher when it is an option and not anobligation for the firm concerned to perform its activities externally.8 Inother words, mandatory subcontracting, which is an induced practicearising from regulation that restricts a firm’s supply chain, tends to in-crease procurement costs. We argue that these results are explained by the

interplay of the following factors. If a firm can choose to subcontract, itwill do so only if subcontracting is profitable. In this case, subcontractingalso implies that firms outsource part of the work to “similar and known”

firms and this entails lower search costs, as well as greater informationsymmetry concerning the execution costs. Moreover, being able to choosewhether to outsource some of the works and having greater informationsymmetry together generate for the contractor stronger bargaining powerin optional than in mandatory subcontracting transactions.

Related Literature. This article provides three main contributions. First,it contributes to the recent empirical literature on subcontracting in publicprocurement that has investigated which rules and/or factors lead to bothcost efficiency and quality in the execution of outsourced contracts. Asalready discussed, De Silva et al. (2012) and Marion (2009) focus on theprocurement costs of programs supporting subcontracting in favor ofspecific groups of firms. Miller (2014) considers the impact of the com-plexity and incompleteness of construction contracts on subcontracting

versus in-house arrangements. Gil and Marion (2013) examine the effectof past and future relationships between contractors and subcontractorsin enforcing informal agreements and the effect of these relationships onthe firms’ bid and entry decisions in the procurement of California high-way contracts. Marion (2012) estimates the effect of horizontal subcon-tracting on bidding strategies in auctions conducted by the CaliforniaDepartment of Transportation and highlights the factors that lead tothis type of outsourcing. As an original contribution to this literature,we add an empirical analysis of procurement price and performance in asetting where rules from supplier qualification screenings affect the firms’subcontracting positions and thus the available alternatives regardingmake-or-subcontract decisions.

Second, our article contributes to the extensive economics literature onfirm boundaries originating from Williamson’s (1971) influential work ontransaction costs. In this respect, we specifically offer an empirical test forthe effect of horizontal and vertical sourcing on the firm bidding for public

8. In a competitive stochastic investment game, Van Mieghem (1999) investigates the

firms’ choice of subcontracting as an option value and finds that the choice of subcontracting

improves the firms’ financial performance and investment coordination under a high degree

of uncertainty.


ownloaded from



I user on 21 Decem

ber 2020

works contracts.9 “Horizontal” subcontracting is defined in Spiegel (1993)as an agreement between rival firms where “each of which is capable ofproducing and marketing its product independently”, while “vertical”subcontracting is an agreement between firms with complementarycapabilities/assets that aims to obtain an output (Webster et al. 1997).In our setting, optional subcontracting can be considered a horizontalagreement, as it occurs between similarly qualified firms, while mandatorysubcontracting can be considered a vertical agreement, as it correspondsto an agreement between firms with different qualifications/capabilities.Comparing bids by firms in optional or mandatory subcontracting pos-itions, we find that optional subcontracting promotes productive effi-ciency. Note that our results are consistent with Lafontaine and Slade’s(2008) findings from an empirical survey on vertical restraints. These au-thors conclude that, when manufactures choose to impose vertical re-straints, their impact on market performance is positive by implicationand if vertical restraints are prohibited, then the impact is negative. Ourempirical findings on subcontracting and Lafontaine and Slade’s findingson vertical restraints both highlight the fact that, on the one hand, thefirms’ voluntary arrangements tend to improve market performance but,on the other hand, any imposed arrangements that either prohibit ormandate relationships tend to worsen market performance.

Third, our article provides empirical evidence regarding the cost arisingfrom procurement rules that limits the discretion of the agents (both thesupplier and the contracting authority [henceforth, CA]) in procurementtransactions. There is a lively debate about the optimal regulation ofpublic contracts through the use of more rigid or more flexible rules af-fecting the procurement transaction outcomes. Spagnolo (2012) high-lighted that this debate directly refers to the inclusion of reputationalforces and arises from two opposite positions. In Europe, regulation hasconstrained the use of past performance information to select contractors,while in the United States, the establishment of databases to evaluate thecompanies’ past performance in public contracts and to share this infor-mation has been encouraged. Specifically referring to the framework forsupplier qualification screenings, this article adds an empirical test for thecost of a planning constraint (i.e., mandatory subcontracting) in the firms’strategy regarding a contract’s execution. Future research is needed toassess the cost of both flexible and rigid rules in procurement, and todesign an appropriate balanced mechanism.

The structure of the article. The remainder of this article is organized asfollows. Section 2 describes the institutional features of public procure-ment auctions and subcontracting in Italy. Section 3 gives detailed infor-mation about the datasets on which we base our investigations. Section 4presents the econometric model, empirical results, and robustness tests,

9. Theoretical contributions on horizontal subcontracting have been provided byKamien

et al. (1989), Gale et al. (2000), and Spiegel (1993).


ownloaded from



I user on 21 Decem

ber 2020

considering all of the firms’ bids. Section 5 illustrates the results of ourestimations when we focus on the bids offered by the winning firms; thatis, the firms that won and fulfilled the contracts. Section 6 presents ourfindings on the measures of ex-post performance. Conclusive commentsare given in Section 7.

2. Regulation on Entry into the Italian Procurement of Public Works

In this section we briefly review the regulation on the supply and demandsides of the Italian market for the procurement of public works, focusingon the rules that directly affect the firms’ subcontracting positions. Wealso provide a brief description of the adopted auction format.

The supply side. According to Italian law on public procurement, firmsmust prequalify to bid in auctions for public work contracts worth morethan 150,000 euros.10 The Italian system for qualifying firms is operatedby a third party (i.e., 37 private companies, called “Societa’ Organismi diAttestazione” [SOA]) that is accredited and monitored by the authority incharge of regulating the national market for public works, supplies, andservices (“Autorita’ di Vigilanza sui Contratti Pubblici di Servizi, Lavori eForniture” [AVCP]11). The firms’ qualifications are provided after estab-lished “general” and “technical” attributes have been ascertained by aSOA. The general attributes concern the firms’ financial standings andcriminal records (e.g., anti-Mafia) and these attributes are the same forany firm wishing to participate in an auction for the procurement of publicworks. The technical attributes have to do with the specific skills that areneeded to perform certain works and are usually assessed based on thefirms’ documented expertise and other observable items. Specifically, theItalian procurement regulation defines 46 “categories of work” underwhich firms can qualify for consideration for a period of 3 or 5 yearsbefore their qualifications must be renewed.

The costs of a firm’s choice to get qualified to enter the market for publicworks can be decomposed into i) an investment cost to gain the capabil-ities necessary for each specific qualification and ii) an administrative costto comply with the required procedures. The latter cost, which consists ofa fee (resulting decreasing in the number of categories of work that thefirm qualifies for), some formal documents, and the time taken to carryout these procedures, is low compared to the advantages for the firm whenit repeatedly wins public contracts.12 The former cost, which refers to theacquisition of firm technologies, capabilities, and expertise required toqualify for each category of work in the public procurement process, islarge and requires specific medium and long-term investments.

10. See: Italian Law No. 163/2006.

11. Since August 2014, AVCP is part of the Italian National Anti-Corruption Authority

(ANAC).

12. Note that the certification body (SOA) charges an extra fee if a qualified firm asks to

delete a previously obtained qualification or add a new qualification.


ownloaded from



I user on 21 Decem

ber 2020

According to the stakeholders, qualified firms tend to maintain the samequalifications from year to year in order to maximize the probability ofrecovering their investment costs by becoming a main contractor, subcon-tractor, or member of a consortium in the execution of public works.

The demand side. A large part of the Italian public procurement marketconsists of contracts that are typically awarded by local CAs (i.e., munici-palities, provinces, regions).13 When awarding a contract, the CA shouldspecify all the tasks (the categories of work) and distinguish between themain and secondary categories of work that are included in the project. Forexample, consider a contract for the building of a road in a new residentialarea. The fulfillment of this contract involves three tasks: tA (road works),tB (water works), and tC (sewage works). Accordingly, in the call for ten-ders, the CA will present task tA as the main category of work and theremaining two tasks (tB and tC) as the secondary categories of work. Thisdistinction is relevant, as participation in the tender process is restricted tofirms that are qualified for the main work category. Alternatively, firmsthat are not qualified for the main work category can participate in theauction as part of temporary consortia (called “Associazioni Temporaneed’Impresa” [ATI]). These consortia are created ad hoc to bid for a givencontract and involve at least one firm that is fully qualified for the maintask.14

For the secondary categories of work involved in a public contract upfor tender, the bidding firm may either be qualified or not. If a firm isqualified, the firm winning the contract can choose either to complete allof the works on its own or to subcontract parts of the works to othersimilarly qualified suppliers (i.e., giving rise to optional subcontracting).On the other hand, if a firm is not qualified for one (or more) of thesecondary categories, it can still bid for the contract, but it will have tosubcontract the works, where it lacks the necessary qualifications, to otherqualified firms (i.e., mandatory subcontracting).15

Considering optional and mandatory subcontracting in terms of firmintegration, the former type of subcontracting can be interpreted as hori-zontal outsourcing because it occurs between two similarly qualified firms.

13. According to the 2013 AVCP Annual Report, approximately 63% of contracts in

2011, with a value of 1160 million euros, were awarded by local governments.

14. We can reasonably assume that consortia bidding for tendered contracts are qualified

to perform all the categories of work involved in a project, as each consortium is established

ad hoc for a tendered contract. Note that in our dataset, bidding behavior is not (statistically)

different between consortia and firms in an optional subcontracting position; this evidence

holds even when auction and firm characteristics are controlled for.

15. As a remote alternative, the firm can lease the qualification for a task for which it lacks

qualifications from a qualified firm that is not bidding for the contract. This is a rarely used

practice because it entails a very expensive agreement (called “avvalimento”). Note that the

supplier can also subcontract part of themain category of work but cannot subcontract more

than 30% of the category’s value. This is a not widely adopted practice and, according to our

definition, it corresponds to optional subcontracting, as the bidder should be qualified for the

main task in order to participate in the auction.


ownloaded from



I user on 21 Decem

ber 2020

The latter type of subcontracting corresponds to the required vertical

outsourcing because it occurs between two firms with complementary

capabilities.The auction format. In the Italian procurement of public works, quali-

fied firms participate by offering a bid consisting of a percentage reduc-

tion—a rebate—on the reserve price set by the CA. The most widely

adopted auction format is based on an average bid auction (ABA) in

which the winner is determined as follows. Given the distribution of all

bids received in an auction, a first average (A1) is computed by averaging

all the bids, except those located in the first and last deciles. A second

average (A2) is then computed by averaging all of the bids above A1

(again, excluding those bids located in the last decile). The winning bid

is the one immediately below A2 (see Figure B1, upper panel, in Appendix

B for an illustration).16

The essential theoretical prediction of a general ABA format (where all

bids are considered and the one closest to the average wins) is that all

bidders have an incentive to submit identical bids, leading to a potential

continuum of Nash equilibria in which all firms submit the same bid at a

price that is high enough to ensure profitability for all bidders. Empirical

and theoretical research on ABA indicates that prices should be higher

than those found using a first-price format and that the random selection

of a winner could diminish performance (see Albano et al. 2006, and

Decarolis 2013). Recent experimental research comparing ABA and

first-price auctions indicates that bids in both formats are “strongly influ-

enced by cost signal” and that “there is no statistically significant differ-

ence in bidding behavior” between the two formats (Chang et al. 2013: 13

and 11, respectively). This experimental evidence on bidding behavior,

coupled with the ABA rule that awards the contract to the (somehow

defined) average bid, seems to support the CAs’ repeated adoption of

ABA in the procurement of public works. In some countries, this is

deemed to be a reasonable mechanism to adequately compensate suppliers

and avoid supplier defaults.Summary. The aim of the Italian system for firms’ qualifications is to

restrict auction participation to firms that are capable of executing the

contract’s main category of work with a reasonable degree of certainty.

For the secondary categories of work involved in a tendered project, firms

may or may not be fully qualified. In the latter case, firms are obliged to

subcontract the work to qualified firms, as all aspects of the project must

be handled by qualified firms.

16. In the awarding phase, bidding firms observe a reserve price that is computed by the

CA using various formulas that typically overestimate contractual costs, thus resulting in an

ineffective binding. Note also, if there are fewer than 10 bidders the lowest and highest bids

are not considered when computing the first average (A1); if there are fewer than five bidders,

the project is awarded to the firm that has offered the highest rebate. In our sample of 269

auctions, only one auction has fewer than five bidders and nine have fewer than ten bidders.


ownloaded from



I user on 21 Decem

ber 2020

A noteworthy direct consequence of this public procurement regula-tion is that, when a contract is tendered, the potential position of eachbidder concerning any subcontracting is well defined, as the categories forwhich the bidding firms should be qualified are also announced. Thismeans that each firm bidding for a public contract is aware that, if itwins, it may outsource some of the tasks for which it is fully qualifiedor it will be obliged to subcontract certain tasks for which it is not qua-lified. Thus, the regulation on bidder qualifications for public contractsallows firms to assess their own production strategy at the bidding stage.This process permits us to observe the two subcontracting formats and,accordingly, the firms’ bids. For the sake of our analysis, it is important tobear in mind that, within the framework we are investigating, the samebidder may be in a position to adopt optional subcontracting for someauctioned contracts, but mandated subcontracting for some othercontracts.

3. Data

Different data sources were used to assemble our original dataset for thepurposes of the present analysis. Detailed information on each open ten-dered public contract has been taken from a hitherto unexploited dataset,consisting of the transcripts of competitive auctions conducted from 2000to 2008 by the Regional Government of Valle d’Aosta (the CA).17 Eachtranscript contains information on the auction ID, number of bidders,bidders’ names, and bids. The auction ID enabled us to access other de-tails of the tendered contracts from a national dataset managed by theAVCP, containing all the contracts with a reserve price higher than150,000 euros. This dataset includes information on the contract awardingprocedure, the reserve price of the contract, the categories of workinvolved in the contract, and the final price paid by the CA and thetiming for the completion of the project.

Information on the suppliers’ qualifications was extracted from anothernational AVCP dataset known as the “Casellario SOA”, a kind of na-tional register that collects the qualification status of every firm for eachwork category.

In summary, for each tendered contract, we have information on thequalifications that are required to complete the tasks involved in the con-tract and all of the actual qualifications held by each bidding firm.Matching these data enabled us to disentangle the firm bids in an optionalsubcontracting position that would have the choice to subcontract fromthe firm bids that would be obliged to proceed with a mandatory subcon-tract for part of the project.

17. Valle d’Aosta is a small mountainous region (3263 sq km, 951MSL) with a population

of 129,000 on Italy’s northwestern border with France and Switzerland.


ownloaded from



I user on 21 Decem

ber 2020

3.1 Descriptive Statistics

Our dataset consists of public contracts that were awarded by the

Regional Government of Valle d’Aosta by means of open tenders.18

Once the CA acquired the bidders’ qualifications regarding legal, fiscal,

economic, financial and technical requirements, the contract was awarded

according to the ABA. In the setting we investigated, this auction format

was used for 89.2% of the auctions in our sample. For the remaining

10.8%, a similar average bid mechanism was combined with a type of

lottery to determine the winning bid.19

Our dataset covered 269 auctions for public contracts, where a total of

13,331 prices were offered by bidders consisting of 892 firms and 1777

temporary consortia.20 The average reserve price was approximately 1.1

million euros (ranging from 156,000 to 5.3 million euros). In terms of the

various tasks, these contracts represented mainly road works (37.2%),

river and hydraulic works (29.7%), and buildings (14.9%); see Table 1

where further summary statistics are also included.As shown in Table 2, 73.8% of the bids in our sample were offered by

firms that had all the necessary qualifications and could therefore opt to

horizontally subcontract part of the work, while 12.9% of the bids were

offered by firms that were not qualified for some of the secondary cate-

gories of work and would therefore be obliged to subcontract some work

to other qualified firms. Finally, approximately 13.3% of the bids were

offered by consortia.In our sample each firm participated in about five auctions (13 auctions

if we exclude consortia), on average. As previously noted in Section 2

(about the supply side), firms tend to maintain their qualifications from

one year to the next, leading to different firm subcontracting positions in

different contracts. Given each firm’s qualifications, we observe that each

firm’s subcontracting position varies according to the tasks included in the

18. According to EU directives, public procurement in Italy can take place through four

types of awarding procedures: open, restricted, negotiated, and competitive dialogue. In our

study, we consider only those cases involving open tenders (the so-called “pubblico incanto”).

Participants in restricted and negotiated tenders are invited by the CA, so including such cases

in our analysis may bias our results because the CA could invite firms with particular features

and qualifications. We have no data concerning contracts awarded using competitive dia-

logue procedures.

19. This auction format works as follows: Given A2 computed as above, a random integer

R between 1 and 9 is extracted. The R-th number among the nine equidistant numbers

between the bid just above the first decile and the bid just below A2 is averaged with A2 to

obtain the winning threshold. The winning bid is the bid immediately above this winning

threshold. We define A3 as the winning threshold that results when R¼ 1 (see Figure B1,

lower panel, in Appendix B). As shown in Galavotti et al. (2014), the mean rebate is lower in

the ABA format that is combined with a type of lottery than in the other ABA format;

however, the bidding behavior is similar in both formats.

20. Note that the total number of original bids for these 269 auctions was approximately

10% higher. However, for this share of bids, we do not have information on the bidders’

qualifications, so these bids were not considered in our final sample of 13,331 bids.


ownloaded from



I user on 21 Decem

ber 2020

contract. In our sample, approximately 74.7% of the bids were offered byfirms that took part both in auctions where they had an optional subcon-tracting position and in auctions where they had a mandatory subcon-tracting position. Approximately 23.9% of the bids were offered by firms(including consortia) that always had an optional subcontracting statusand approximately 1.4% of the bids were by firms that were always com-mitted to adopting mandatory subcontracting.

The descriptive statistics for our sample gives us a clear idea of the localdimensions of the market for procuring public works in Valle d’Aosta(see Table 2). Approximately 35.3% of the participants in the auctions(corresponding to 32.4% of the bids) were firms located in the region, and22.6% of the participants (27.6% of the bids) were firms from the larger

Table 2. Descriptive Statistics: Bidders’ Characteristics

Percentage

Local bidders (% of bids) 32.37

Bidders’ size (% of bids):

small 11.80

medium 52.86

large and co-operatives 22.01

Consortia (% of bids¼% of bidders) 13.33

Subcontracting status (% of bids):

Mandatory 12.86

Optional (excluding consortia) 73.81

Subcontracting status (% of bids):

Always mandatory firms 1.40

Sometime optional and sometime mandatory firms 74.72

Always optional firms (excluding consortia) 10.55

Note: See Appendix A for the definitions of the variables.

Table 1. Descriptive Statistics: Bidding Offers and Contract Characteristics

Procurement projects issued by Valle d’Aosta Regional Government

Variable Obs. Mean St.Dev. Min Max

Bid (Rebate, in %) 13331 17.210 4.831 0.001 43

Reserve price (euros) 269 1103786 865298.5 155526.3 5267860

No. of participants 269 55.450 31.845 3 155

Expected duration (days) 269 311 168.179 79 1440

ABA 269 0.892 0.311 0 1

ABA + lottery 269 0.108 0.311 0 1

Road works 269 0.372 0.484 0 1

River and hydraulic works 269 0.297 0.458 0 1

Building 269 0.149 0.356 0 1



ownloaded from



I user on 21 Decem

ber 2020

neighboring Piedmont region. The remaining 42.1% of participants(40.0% of the bids) came from other parts of Italy. The average distancebetween the bidder’s location (i.e., the closest capital city) and the CAlocation (i.e., the city of Aosta) was approximately 310 km (with a stand-ard deviation of 399 km). In terms of the prices offered, local firms offeredan average discount of 17.0%, which was slightly lower than the meandiscount of 17.3% by non-local firms. Similarly, the backlog of projects(i.e., the number of ongoing public projects each firm has at the time ofbidding) was lower for the local firms (1.1 projects) than for the other firms(2.8 projects).21

4. Analysis of the Bidding Offers

4.1 Testable Hypothesis and Model Specification

In this section, we consider the bids offered by all of the participating firmsin the 269 auctions for the procurement of public contracts. We considerthe bid as a proxy for the value attributed by the firm to the project; that is,what the firm expects the work to cost, plus a mark-up. We reference ourapproach to the recent experimental findings by Chang et al. (2013), whocompared the first-price (i.e., “low bid”) and “average bid” methods, andfound that bidding behavior is identical in the two formats and stronglyinfluenced by cost signals. Accordingly, we simply assume in this studythat the bid is a proxy for each firm’s expected costs to complete thetendered contract.22

Mandatory and optional subcontracting in this procurement setting canbe considered a planning constraint and a planning alternative, respect-ively, in the firms’ strategy regarding the contract’s execution, and conse-quently, with different effects on the expected cost and, therefore, thefirms’ bids. Our testable hypothesis is that there may be a significant dif-ference between firms who are obliged to subcontract and firms who are ina position to complete the contracted work themselves or outsource partof the work.

A simple two-group mean-comparison test (Table 3) shows that theaverage rebates offered by firms obliged to subcontract are significantlylower (i.e., corresponding to higher prices to be paid by the CA) than thoseoffered by firms with the option to subcontract. The picture does notchange after we exclude the consortia from the sample.

21. The distance is defined as the number of kilometers between the capital city of the

supplier’s province and the city of Aosta (i.e., the capital city of the single-province region of

Valle d’Aosta). The firm backlog is defined as the number of pending projects (also consider-

ing the public procurement works tendered in the other Italian regions) the firm had at the

date of bidding. Because we do not have data on firm size, we use the type of business entity as

a proxy; approximately 75% of the bids were offered by medium and large firms. See

Appendix A for detailed definitions of the variables.

22. Bajari et al. (2014) empirically investigate highway procurement in the State of

California and show that adaptation costs are an important determinant of the firms’ bids

in public procurement, accounting for 8–14% of the winning bids.


ownloaded from



I user on 21 Decem

ber 2020

These results may be due to various factors associated with the charac-teristics of the firms and auctions concerned. For instance, larger and/ormore efficient firms may be more likely to be fully qualified because theyare qualified for a larger number of categories of work. To check for theinfluence of these factors and to examine the differences in the priceoffered by bidders in the two subcontracting positions, we estimate thefollowing model specification for bidding rebates:

Rebateij ¼ �+ �Optionalij + �Qj + �Xi + lZij + eij ð1Þ

where Optional is a dummy variable with a value of 1 when firm i is fullyqualified to handle project j’s tasks (i.e., in an optional subcontractingposition) and a value of 0 when the firm is partially qualified (i.e., in amandatory subcontracting position). Qj is a set of variables that controlsfor the nature of the project and auction (i.e., proxies for the characteris-tics of the project, such as its dimension or complexity and type of workinvolved; proxies for the characteristics of the auction, such as the formatand the level of competitive pressure; and year dummy variables to adjustfor temporal shocks that may have affected both the time-related trends ofthe firms’ bidding behavior and the contracts chosen by the CA). Xi rep-resents a set of firm characteristic features (i.e., proxies for firm size andthe distance between the firm location and CA). Zij controls for the firm’scapacity constraints, which is proxied by a measure of the firm’s backlogwhen it bids for a project, and "ij is the error component.

In some model specifications, to reduce the omitted variable problems,we include firm-fixed effects to adjust for firm-specific characteristics(e.g., size, productivity, financial position, and location). This enables usto focus on the within-firm variation in optional or mandatory subcon-tracting status and to capture the effect of changes therein better. Thesefirm-specific characteristics could also vary over time, so in different spe-cifications of the model we include firm-year-fixed effects to focus on thewithin-firm-year variations in the subcontracting positions.

4.2 Estimation Results

Our primary coefficient of interest is �, which indicates whether a firm’ssubcontracting status affects its bidding offer. This coefficient reflects the

Table 3. Correlation: Subcontracting Status and Bidding Offers

Mean rebate Mean rebate (excluding consortia)

Optional 17.348 17.339

Mandatory 16.272 16.272

Difference 1.076*** 1.067***


Inference: *** p<0.01, ** p< 0.05, * p< 0.1.


ownloaded from



I user on 21 Decem

ber 2020

difference between the rebates offered by firms that can choose to subcon-

tract and the rebates offered by those obliged to subcontract. To addresspotential heteroscedasticity issues, we use ordinary least squares (OLS)

estimates with robust standard errors clustered at the firm level to enable

correlations among within-firm observations.Our results are presented in Table 4, columns 1–3, and show that the

coefficient for the Optional variable always has a positive sign and is stat-

istically significant.23 All else remaining equal, firms that are fully qualifiedto complete a project and, thus, may or may not subcontract part of the

work, offer significantly greater rebates (i.e., corresponding to lower

prices) than firms that are obliged to subcontract part of the work to

other qualified firms. In particular, fully qualified firms offer discountsapproximately 0.3 percentage points larger than partially qualified firms.

This value corresponds to about 2% of the average bidding discount

observed in our sample.These findings show that a bidding firm’s production efficiency in-

creases (i.e., via lower production costs) when its subcontracting position

is flexible and this enables the firm to subcontract part of the project tosimilar firms. This is not the case when firms are required to subcontract

part of the work to other complementary firms.We interpret our findings from the following considerations. If a firm

can choose to subcontract, it will do so only if subcontracting is profit-

able.24 Optional subcontracting implies that firms outsource part of thework to “similar and known” firms; thus, optional subcontracting is asso-

ciated with lower search costs, as well as greater information symmetry

concerning the execution costs, than any form of required subcontracting

to firms with “different and complementary” qualifications. Moreover,

being able to choose whether to subcontract part of the work and tohave greater information symmetry combined to give a firm stronger bar-

gaining power in optional compared to mandatory subcontracting.These considerations are based on several studies of the effect of the

parties’ information asymmetry and bargaining power on the outsourcing

choice. For the effects of information asymmetry, Lewis and Sappington(1991) use a standard procurement model and assume that firms are better

able to monitor effort in their internal tasks rather than in outsourced

tasks. Because the subcontractor may have a lower cost technology, the

23. Note that, in Table 4, column 1, we estimate our full sample of bidding rebates; in

column 2 we include firm-fixed effects and focus on the bids offered by firms that we observe

more than once and took part both in auctions where they had an optional subcontracting

position and in auctions where they had a mandatory subcontracting position; while, in

column 3, we include firm-year-fixed effects and focus on the bids offered by firms that we

observe more than once in a given year and took part both in auctions where they had an

optional subcontracting position and in auctions where they had amandatory subcontracting

position in a given year.

24. Quinn and Hilmer (1994) present an extended discussion on the firms’ relative risks/

costs and benefits from outsourcing in different industrial sectors.


ownloaded from



I user on 21 Decem

ber 2020

authors argue that the decision to outsource involves a trade-off betweenlower production costs and higher monitoring costs. Thus, the supplieropts to use subcontractors when the efficiency gains from lower produc-tion costs are greater than the loss of control resulting from the highermonitoring costs of outsourcing. In our setting, monitoring costs arelower for optional subcontracting than for mandatory subcontracting,as the work with optional subcontracting is delegated to similarly qualifiedfirms.25

For the effects of bargaining power, Grossman and Helpman (2002)investigate how the distribution of bargaining power between the subcon-tractor and the supplier affects the viability of outsourcing. Their modelshows that a generic task—as opposed to a specific one—enhances outsideoptions, which improves the task producer’s bargaining power. Similarly,

Table 4. Estimation Results: Bidding Offers

Dependent variable: Bidding rebate

Mean outcome: 17.210 17.151 17.530

OLS regressions

1 2 3

Optional 0.224** 0.316*** 0.403***

(0.099) (0.100) (0.109)

log(Reserve price) 0.104 0.143* 0.122

(0.065) (0.076) (0.088)

log(Expected duration) �0.097 �0.093 �0.235*

(0.084) (0.094) (0.128)

log(No. participants) 1.242*** 1.340*** 1.234***

(0.137) (0.153) (0.199)

log(1 + Backlog) �0.065 0.035 �0.225

(0.063) (0.085) (0.213)

log(Distance) 0.007

(0.029)

Category of work dummy YES YES YES

Type of auction dummy YES YES YES

Firm size dummy YES NO NO

Firm-fixed effects NO YES NO

Firm-year-fixed effects NO NO YES

Year dummy YES YES NO

Observations 13,331 9961 6280

R-squared 0.520 0.562 0.575

Note: See Appendix A for the definition of the variables.

Robust standard errors clustered at firm-level in parentheses.

Inference: ***p< 0.01, **p< 0.05, *p< 0.1.

25. Riordan and Sappington (1987) provide a two-stage production model in which costs

are observable only by the producing party. The information asymmetry between parties and

the correlation between first- and second-stage costs determine the principal’s optimal choice

in the organization of task productions.


ownloaded from



I user on 21 Decem

ber 2020

in our setting, a firm that can choose to delegate part of the work (i.e.,optional subcontracting) is endowed with an outside option that enhancesits bargaining power in outsourcing. This outside option is not present inthe case of mandatory subcontracting.

Our results are also consistent with the study by Lafontaine and Slade(2008) regarding firm boundaries and the effects of rules. In their setting,the authors find that voluntary arrangements among firms have a positiveimpact on market performance (i.e., lowering both firm costs and con-sumer prices), while government-mandated arrangements systematicallyhave a negative impact on market performance.

For other firm characteristics, the model specification in column 1 ofTable 4 includes dummy variables for firm size, the distance between thefirm’s location and CA, and a measure of the firm’s backlog. The model incolumn 2 includes dummy variables for firm-fixed effects, which enables usto control for firm characteristics that do not vary over time. Moreover,the model in column 3 includes dummy variables for firm-year-fixedeffects, which aim to capture a firm’s characteristics (e.g., its size, financialposition, and productivity) in any given year. The use of firm-fixed effectsin the model also allows us to exclude from our sample consortia and firmsthat always or never had all of the necessary qualifications, and thus, toonly concentrate on the bidding firms that were fully qualified for someauctions and partially qualified for others. Concentrating on such firms isimportant for two reasons. First, to avoid any biases in our estimates thatmight have stemmed from the inclusion of consortia in our sample (withthe corresponding assumption regarding that these consortia had all thenecessary qualifications). And second, and more importantly, to supportour inference that the overall results are not influenced by firms thatalways or never had all of the required qualifications. This approachallows us to exploit the within-firm (or within-firm-year) variations insubcontracting status.

Our estimates of the other control variables in the model specificationsare comparable to the results obtained in previous empirical studies on theawarding of public procurement contracts. It is hardly surprising thatrebates are positively influenced by the size of a project and the numberof participants, and negatively influenced by the expected duration of thework (the size and duration measures are both calculated by the CA en-gineers and are known by the firms before they place their bids).26

26. In the United States, Bajari et al. (2009) show that having more firms competing in an

auction reduces bidding prices. Their finding is based on a dataset of contracts awarded by

private authorities in the building construction industry in northern California from 1995 to

2001. Similarly, based on a sample of Italian public procurement auctions, Bucciol et al.

(2013) finds that a larger number of bidders increases the amount of the winning bidder’s

rebate. Our results confirm the positive (although weakly significant) effect of the reserve

price on the rebates offered, as also reported by Coviello and Gagliarducci (2010) and

Decarolis (2009). Note that, in our estimates, the results for the controls for backlog and

distance between the bidder and CA are not statistically significant.


ownloaded from



I user on 21 Decem

ber 2020

4.3 Robustness Checks

A first concern regarding our estimates has to do with the influence ofextreme bids. In fact, it may be that outlying bids drive the estimation ofthe coefficient for our Optional variable of interest. We address this con-cern by using a robust regression approach (iteratively reweighted leastsquares, [IRLS]) that iteratively assigns a lower weight to outlying obser-vations. As shown in Table 5, column 1, the estimated coefficient of theoptional subcontracting status is positive and statistically significant, thusconfirming the previous estimates.

A further concern is related to the likelihood that the estimated differ-ence in offered bids between firms in optional subcontracting positionsand firms in mandatory subcontracting positions is driven by very differ-ent rebate distributions across auctions. In fact, the number of partici-pants and their bids vary across auctions, resulting in a different rebatedistribution for each auction. Even if we control for several characteristicsof the auction in the model specification, we may not fully capture thesedifferences.

The average bid auction adopted in order to award the public procure-ment contracts enables us to identify different areas in the distribution ofthe bidding firms’ rebates. In particular, we distinguish an area around thewinning rebate as follows. The winning offer in each average bid auction isbetween the mean rebate (A1) and the rebate corresponding to the 90thpercentile of the distribution (“area A”, see Figure B1, upper panel, inAppendix B). In the average bid auctions combined with a lottery, thewinning offer is between lowest possible winning threshold (A3) and therebate at the 90th percentile of the distribution (“area A”, see Figure B1,lower panel, in Appendix B). We focus on the “area A” of the auction-specific distribution of the rebates and check whether the previously esti-mated difference holds. We perform this check because one might suspectthat our results are driven by the fact that bidders subject to mandatorysubcontracting tend to offer particularly small rebates; that is, those bidson the left-hand side of the distribution. If this were true, subcontractingstatus may not be the only difference between the two types of bidders.There may be other differences related to the firms’ productivity and tech-nology. Moreover, firms bidding in a mandatory subcontracting positionwould not be competitive enough to win the auction and they may takepart in auctions for collusive purposes; that is, to favor a given bidder (or agroup of bidders).

Therefore, after ensuring that the subcontracting position is not a sig-nificant determinant of the bidders’ likelihood to offer a bid in “area A” ofeach auction distribution,27 we concentrate our analysis on “area A” ofeach auction distribution and we nonetheless find that the difference in the

27. Note that in terms of distance from the winning bid, firms in both optional and

mandatory subcontracting positions offer, on average, rebates lower than the winning

rebate (�0.812 and �0.617 percentage points, respectively).


ownloaded from



I user on 21 Decem

ber 2020

rebates offered between firms in the two different subcontracting positionspersists and is statistically significant (Table 5, column 2).

5. Analysis of the Winning Offers

This section analyzes whether the more aggressive bids offered by firmsthat can choose whether or not to subcontract result from the expectedpotential advantage of outsourcing or the reluctance of such firms to usesubcontractors. To ascertain whether the firms’ optional subcontractingstatus is associated with lower prices, we examine the winning firms thatcould opt for outsourcing and actually used subcontractors to completepart of the awarded projects.

With this aim, we now consider two samples: one consists of 220 win-ning bids drawn from the sample of auctions held by the RegionalGovernment of Valle d’Aosta;28 the other (which served to test the ro-bustness of our estimates) includes winning bids in 449 auctions held byseveral CAs in Valle d’Aosta between 2000 and 2009. For the latter sampleof auctions, we only know the characteristics of the winning firmsand their rebates; we have no information on the bids of the other

Table 5. Robustness Checks: Bidding Offers

Dependent variable: Bidding rebate

Mean outcome: 17.210 17.919

Robust reg. OLS

Sample: All bids Bids in “Area A”

1 2

Optional 0.149*** 0.480***

(0.043) (0.096)

Res.price; Exp.dur.; No.part.; Backlog YES YES

Category of work dummy YES YES

Type of auction dummy YES YES

Firm size; Distance YES NO

Firm-fixed effects NO YES

Year dummy YES YES

Observations 13,331 3798

R-squared 0.874 0.766


Standard errors are in parentheses. In column 2, robust standard errors are clustered at the firm-level.

Inference: ***p< 0.01, **p< 0.05, *p< 0.1.

28. This sample is extracted from the 269 auctions held by the Regional Government of

Valle d’Aosta. For 49 of these auctions, we do not have full details regarding the number of

subcontractors and the value of the subcontracted work, or we do not have information to

determine the subcontracting position of the winning firm. See the upper panel in Table 6 for

the summary statistics for this sample.


ownloaded from



I user on 21 Decem

ber 2020

participants.29 For each project, we obtain information from the AVCPdataset on the extent of subcontracting and the number (and ID) of thesubcontractors that were used by the winning firm (see the summary stat-istics in Table 6).

The descriptive statistics for the sample of winning rebates in theRegional Government’s auctions indicate that approximately 85.0% ofsuppliers actually subcontracted at least part of the work and approxi-mately 88.2% of the firms had an optional subcontracting status. Alsonote that, excluding consortia, firms in optional subcontracting positionsthat did subcontract outsourced a smaller proportion of their project’svalue (approximately 259,000 euros, on average) than firms in mandatorysubcontracting positions (approximately 275,000 euros, on average).Similarly, winning firms that had the option to subcontract outsourcedto a smaller number of subcontractors (1.8) than firms that were requiredto outsource (2.5), as well as the firms belonging to a consortium (2.1).

5.1 Estimation Results

To study the rebates offered by winning firms that actually engage insubcontracting, in our model specification, we include the interaction be-tween the firm’s optional subcontracting status and the variable indicatingthe firm’s actual use of subcontracting (Sub*Optional). Because Italianregulation prohibits a firm in a mandatory subcontracting position fromnot subcontracting, it is necessary to exclude (Sub) from the set of dummyvariables in the model.

The results are presented in Table 7 columns 1 and 3, and show that the(Sub*Optional) interaction term is positive and statistically significant,while the (Optional) coefficient is positive (and statistically significant incolumn 3). This indicates that the effect on the winning rebates related to amandatory subcontracting position is negative and that the effect relatedto an optional subcontracting position is especially large when the firmactually subcontracts.30

29. The summary statistics for this sample are presented in the lower panel in Table 6. In

this sample, 58.4% of the projects were for the Regional Government, 34.5% were for

municipalities, and the remainder was for other local public authorities (e.g., territorial as-

sociations for mountainous areas). Note that for the auctions appearing only in this sample,

we are not able to distinguish between ABA and ABA + lottery awarding mechanisms. Note

also that the smaller set of winning offers (220) for contracts awarded by the Regional

Government is a sub-sample of this larger sample (449).

30. To deal with any outliers, we use robust regressions (IRLS), which iteratively assign a

lower weight to deviant observations. The winning prices are generally distributed in the same

way in the two samples; however, when the distribution of the winning prices is compared

with the distribution of all prices offered for contracts with the Regional Government of Valle

d’Aosta, the presence of outlying observations seems to have more weight in the distribution

of the winning price.


ownloaded from



I user on 21 Decem

ber 2020

These empirical findings are in line with Spiegel’s (1993) theoreticalresults on horizontal subcontracting, which indicate that this form of out-sourcing allows firms to improve their production efficiency and increasessocial welfare for a large set of parameters in his model.

In our setting, horizontal subcontracting corresponds to optional sub-contracting and firms engaging in optional subcontracting can indeedoffer higher rebates than firms engaging in mandatory subcontracting (aform of vertical outsourcing). In particular, our estimates indicate that theoption to subcontract induces firms to outsource only when it is profitable.The option to outsource, in turn, gives the firm a stronger bargainingposition than the obligation to outsource part of the project. This conclu-sion is confirmed by the results in Table 7, columns 2 and 4, where theauction samples are restricted to contracts where at least part of the pro-ject was handled by subcontractors (i.e., the focus in these columns islimited to projects involved in subcontracting). Again, the coefficient esti-mated for the Optional variable is positive and statistically significant,indicating that firms in an optional subcontracting position won the

Table 6. Descriptive Statistics: Winning Offers and Contract Characteristics

Procurement projects issued by Valle d’Aosta Regional Government

Variable Obs. Mean St.Dev. Min Max

Winning rebate (%) 220 17.222 4.260 3.620 31.990

Sub 220 0.850 0.358 0 1

Optional 220 0.882 0.324 0 1

No. of subcontractors 220 1.645 1.527 0 11

Value of subcontracts (euros) 220 244543.9 297106.3 0 1800620

Bidder-Subcontractor 220 0.409 0.493 0 1

Reserve price (euros) 220 1137356 894343.5 155526.3 5267860


Expected duration (days) 220 310.145 155.766 79 899

Road works 220 0.341 0.475 0 1


Buildings 220 0.150 0.358 0 1

Procurement projects issued within the borders of Valle d’Aosta by several CAs

Winning rebate (%) 449 16.068 4.717 1.900 36.639

Sub 449 0.788 0.409 0 1

Optional 449 0.768 0.422 0 1

No. of subcontractors 449 1.724 1.977 0 17

Value of subcontracts (euros) 449 213693.2 329375.1 0 3256000

Reserve price (euros) 449 973631.8 884015.4 151457.4 6180000


Expected duration (days) 449 301.987 170.973 59 977

Road works 449 0.336 0.473 0 1

Buildings 449 0.203 0.402 0 1


Note: See Appendix A for the definition of the variables.


ownloaded from



I user on 21 Decem

ber 2020

contract with a discount approximately 1.3 percentage points higher thanfirms in a mandatory subcontracting position. This value corresponds toapproximately 7.5% of the average winning percentage discount observedin our data.

In the estimated model specifications in Table 7, columns 2 and 4, weinclude two further variables—the number of subcontractors (No. subcon-tractors) and the value of the subcontracts (Value of subcontracts)—toinvestigate whether these variables affect the bids offered and influencethe validity of our findings. The estimated coefficient of the variable forthe number of subcontractors is positive and not statistically significant,31

Table 7. Estimation Results: Winning Offers

Dependent variable: Winning rebate

Mean outcome: 17.222 17.183 16.068 16.069

Robust regressions

CA: Regional Government Several CAs

Sample: Full Only sub Full Only sub

1 2 3 4

Optional 0.408 1.278*** 0.627* 1.279***

(0.370) (0.305) (0.321) (0.264)

Sub*Optional 0.764*** 0.500*

(0.260) (0.264)

log (Value of subcontracts) 0.198 0.279*

(0.149) (0.145)

log (No. subcontractors) 0.049 0.027

(0.190) (0.204)

Bidder-subcontractor �0.198

(0.201)

Res.price; Exp.dur.; No.part.; Backlog YES YES YES YES

Category of work dummy YES YES YES YES

Type of auction dummy YES YES NO NO

Type of CA dummy NO NO YES YES

Firm size; Distance YES YES YES YES

Year dummy YES YES YES YES

Observations 220 187 449 354


Standard errors are in parentheses.

Inference: ***p< 0.01, **p< 0.05, *p< 0.1.

31. Differently from the setting investigated by Gil and Marion (2013), we rarely observe

repeated interactions between contractors and subcontractors in the public procurement

market of Valle D’Aosta (on average, they interacted only 1.2 times in a decade). Note

that the environment analyzed in Gil and Marion’s paper is characterized by features that

potentially give rise to relational intertemporal incentives between contractors and subcon-

tractors: the subcontractors do not have to be strictly qualified and should be indicated when

the bid is made; and the contractors benefit from having a larger degree of discretion in the

outsourcing choice.


ownloaded from



I user on 21 Decem

ber 2020

while the estimated coefficient of the variable for the value of subcontractsis positive, but only statistically significant in the sample of auctions issuedby several CAs in the region.32

6. Extensions

Thus far, our results indicate that the subcontracting status is stronglyassociated with the rebates offered. We interpret this finding to indicatethat firms discount their outsourcing position in their bids. Consideringthe procurement agencies’ interest, which focuses also on other perform-ance outcomes aside from gathering competitive bids, it is worthwhile toexamine whether the two subcontracting positions are associated withdifferent ex-post performances. To empirically test whether there is a re-lationship between subcontracting position and ex-post performance, weuse data for the probability of time and cost overrun in the execution of aproject. We define time overruns as the probability of completing theproject after the expected (contracted) deadline and cost overruns as theprobability that the final cost of the project is greater than the winning bid.

A priori, the effect of the subcontracting position on ex-post perform-ance is unclear. On the one hand, we might expect that the more aggressivebidding behavior of optional subcontracting firms who offer higher re-bates would result in worse ex-post performance as suppliers try to recoversome mark-up during the executional phase of the project. On the otherhand, mandatory subcontracting firms may not have sufficiently dis-counted their asymmetric information and lower bargaining power pos-itions within their offers, so they may incur additional costs during thecontract’s execution.

In our data, we observe a high probability of time and cost overruns. Infact, time and cost overruns occurred respectively in 91.8% and in 84.5%of the projects for the Regional Government of Valle d’Aosta, while timeand cost overruns occurred respectively in 93.1% and in 86.6% of projects

32. Note that, in column 2 of Table 7, the coefficient for the Bidder-subcontractor variable

(concerning the presence of at least one subcontractor who also participated as a bidder in the

same auction) is not statistically significant. As recently studied by Marion (2013), when a

supplier outsources part of the work to a subcontractor who has participated in the same

auction (i.e., a rival in the auction) and when there is an ex-ante agreement between the two,

maximizing his effort to win the auctionmay not be optimal for the subcontractor because the

opportunity to work as a subcontractor may be more profitable. It is worth stressing here that

in Italian public procurement, bidders should not commit to engage in a subcontracting

relationship with a particular firm at the auction stage. Indeed, bidders do not have to indicate

the names of the subcontractors in their offers. This setting reduces the likelihood that firms

will make ex-ante agreements and thus, such agreements may not have any influence on their

rivals’ costs.


ownloaded from



I user on 21 Decem

ber 2020

awarded by several CAs in the Valle d’Aosta region. We use data for both

samples of projects to test a relationship between firm’s subcontracting

position and the probability of time and cost overruns.The results in Table 8, columns 1–4, do not indicate a clear relationship

between a firm’s subcontracting position and the probability of time and

cost overrun.33 Firstly, the estimated coefficients for (Optional) and

(Sub*Optional) are not statistically significant. Secondly, the signs of the

coefficients estimated for the sample of auctions issued by the Regional

Government and the sample of auctions issued by several CAs in the

region are mixed.When we focus on the sample of the Regional Government (columns 1

and 2), we find that the predicted probability of time overruns, computed

at the mean values of the other regressors, is the highest for firms in a

mandatory subcontracting position (98.3%). This probability is higher for

firms in an optional subcontracting position that do subcontract (94.3%)

than for firms that do not subcontract (91.4%). Similarly, we find that the

predicted probability of cost overruns at the means is the highest for firms

in a mandatory subcontracting position (97.9%), while this probability is

Table 8. Estimation Results: Ex-Post Performance

Dependent variable: Probability of:

Time

overrun

Cost

overrun

Time

overrun

Cost

overrun

rebate

>Win

Mean outcome: 0.894 0.834 0.915 0.869 0.349

Probit regressions

CA: Regional Gov. Several CAs Regional

Gov.

1 2 3 4 5

Optional �0.757 �0.999 0.097 �0.106 0.028

(0.593) (0.718) (0.307) (0.296) (0.042)

Sub*Optional 0.215 0.113 �0.243 �0.003

(0.417) (0.356) (0.275) (0.245)

Res.price; Exp.dur.; No.part. YES YES YES YES YES

Category of work dummy YES YES YES YES YES

Type of auction dummy YES YES NO NO YES

Type of CA dummy NO NO YES YES NO

Firm size; Backlog; Distance YES YES YES YES YES

Year dummy YES YES YES YES YES

Observations 170 199 355 404 13,331


Robust standard errors clustered at the firm-level are presented in parentheses.

Inference: ***p< 0.01, **p< 0.05, *p< 0.1.

33. In the probit regressions, perfectly predicted observations have been dropped.


ownloaded from



I user on 21 Decem

ber 2020

higher for firms in an optional subcontracting position that do subcon-tract (87.6%) than for firms that do not subcontract (85.2%).34 However,when we focus on the sample that includes also contracts issued by otherCAs (columns 3 and 4), the estimation results show that the optionalsubcontracting firms that do subcontract have the lowest predicted prob-ability of cost and time overruns (computed at the mean values of theother regressors).

This unclear relationship is also confirmed in Table 8, column 5, wherewe estimate the probability of very high rebates, which could be associatedwith a higher risk of holding up the buyer ex-post. The estimation resultsshow that optional subcontracting firms do not have a significantly dif-ferent probability of offering rebates that are higher than the winning bid.In other words, our data do not provide empirical evidence supporting thehypothesis that high rebates are associated with ex-post hold-up in theperformance of the contract.

7. Conclusion

We empirically investigate public procurement in Italy, where the existingregulations on the firms’ pre-qualifications for performing public workscontracts determine two different subcontracting positions: (i) the con-tractor is fully qualified and may choose to either subcontract part of thecontract or to complete the work on his own (optional subcontracting); or(ii) the contractor is partially qualified and is obliged to subcontract partof the work to a qualified firm (mandatory subcontracting).

We examine the two subcontracting positions by analyzing data onthe auctions for public works and the bidders’ characteristics. We findthat bidders in an optional subcontracting position offer lower prices(i.e., higher rebates on the reserve prices set by the contracting authorityawarding the contract) than those in a mandatory subcontractingposition.

Our findings are confirmed by different estimation and robustness tests.In particular, when we focus only on the bids made by auction winnersthat subsequently subcontracted part of the contract, we find that the costsof completing the contracts are lower when firms actually engage inoptional subcontracting than when firms are mandated to engage in

34. Note that, the average predicted probability (not computed at the mean values of the

other regressors) of time overruns for the group of firms in a mandatory subcontracting

position is 90.7%, in a optional subcontracting position that do subcontract is 89.2%, and

in optional subcontracting position that do not subcontract is 90.0%. The average predicted

probability of cost overruns for the group of firms in a mandatory subcontracting position is

95.2%, in a optional subcontracting position that do subcontract is 82.8%, and in optional

subcontracting position that do not subcontract is 75.9%.


ownloaded from



I user on 21 Decem

ber 2020

subcontracting. Moreover, we find no significant differences in contractperformance (i.e., the probability of cost and time overruns) between thetwo subcontracting positions.

We interpret these findings as follows. Having the option to use sub-contractors induces firms to do so only when subcontracting part of thework is profitable and when it puts them in a stronger position regardinginformation asymmetry and bargaining power. Firms obliged by regula-tion to engage in mandatory subcontracting lack these advantages andinclude the higher expected costs of outsourcing in their bids.

Our results are consistent with those from an empirical survey of verti-cal arrangements by Lafontaine and Slade (2008). Both the findings ofthese authors and our empirical results indicate that voluntary arrange-ments tend to improve, while imposed arrangements tend to worsenmarket performance. This has interesting policy implications for publicprocurement, a setting in which pre-qualification rules to prevent low-quality procurement transactions are often adopted. All in all, our resultssuggest that limiting the discretion of firms in terms of their supply-chainchoices represents a cost (i.e., higher price) in the public procurementprocess. A way to reduce this cost according to our context could be todefine relatively broader categories of work for firms’ qualifications. In sodoing the winning firms would be fully qualified more often and thuswould exert their more efficient discretional choices for subcontracting.Contextually, this broader regulatory design should be coupled with boththe adoption of reputational criteria in the awarding of contracts (i.e.,the suppliers’ past performance) and the entitlement of power to theprocurement authorities to disqualify any unsuitable firms (Spagnolo2012). This would counterbalance the potential reduction in qualityassociated with the contractors as a consequence of the more leniantpre-qualification screenings. Future research is still needed to understandthis trade-off.

Appendix A

Variables, Definitions and Abbreviations

Rebate (or percentage of the price reduction or discount). The price cutoffered by participants in an auction, expressed as a percentage of theauction’s reserve price.

Optional. A dummy variable taking a value of 1 if the firm can choosewhether to horizontally subcontract part of the contracted work; this firmis fully qualified to complete the project alone, but it can opt to subcon-tract part of the work to firms with similar qualifications. The dummytakes a value of 0 if the firm is required by law to subcontract part of thework that it is not qualified to perform.


ownloaded from



I user on 21 Decem

ber 2020

Mandatory. A dummy variable taking a value of 1 if the firm is requiredby law to subcontract part of the work; this firm does not have all thequalifications to complete the project, and it is required by law to subcon-tract the work for which it is not qualified to perform to firms with therequired qualifications. The dummy takes a value of 0 if the firm (beingfully qualified to handle the work) can choose whether to subcontract partof the work.

Reserve price. The auction’s starting value (in euros) decided by thecontracting authority [CA]. All the projects considered in this analysishad a reserve price higher than 150,000 euros.

Expected duration. The expected duration of the work (in days), decidedby the CA.

No. of participants. The number of bidders participating in an auction.

Firm size. A set of dummy variables that are used as proxies for the size ofthe bidding firms. Because we do not have data on the number of firms’employees or their total assets, we construct proxies based on the type ofbusiness entity (there is a positive correlation between the type of businessentity and the size of the Italian firms). In particular, our proxies aredefined as: Small (one-man businesses, limited and ordinary partnerships);Medium (limited liability companies); or Large + cooperatives (public cor-porations and cooperatives).

Distance. It is a proxy for the distance between the firm and the project.Because we do not have information for the exact location of the projectand of the firm, it is defined as the distance in kilometers from the capitalcity of the province in which the firm is located and the city of Aosta(which is the capital city of the single-province region of Valle d’Aostaand where the CA is located). We assign a distance of 30 km to firmslocated in Valle d’Aosta.

Backlog. A proxy for the firms’ capacity constraints, defined as thenumber of pending projects (considering also public procurementworks tendered in other Italian regions) the firm has at the date ofbidding.

Consortia. A dummy variable taking a value of 1 when it refers to atemporary association of firms, and 0 otherwise. Firms can join forces,pool their qualifications and form a consortium to participate in a givenauction, so we assume that the variable Optional takes a value of 1 forconsortia.


ownloaded from



I user on 21 Decem

ber 2020

Type of auction. A set of dummy variables describing the auctionmechanism.

ABA is an average bid auction defined as follows. Given the distributionof all bids received in an auction, a first average (A1) is computed byaveraging all bids except those located in the first and last deciles. Asecond average (A2) is then computed by averaging all of the bidsabove A1 (again, excluding those bids located in the last decile). Thewinning bid is the one immediately below A2 (see Figure B1, upperpanel, in Appendix B). If there are fewer than ten bidders, the lowestand highest bid is not considered to compute the first average (A1). Ifthere are fewer than five bidders, the project is awarded to the firm that hasoffered the highest rebate.

ABA+lottery is an average bid auction defined as follows. Given A2computed as above, a random integer R between 1 and 9 is extracted.The R-th number among the nine equidistant numbers between the bidjust above the first decile and the bid just below A2 is averaged withA2 to obtain the winning threshold. The winning bid is the bid imme-diately above this winning threshold. We define A3 as the winningthreshold that results when R¼ 1 (see Figure B1, lower panel, inAppendix B).

Category of work. A set of dummy variables representing the main cate-gory of work in a project (e.g., road works, buildings, hydraulic works,etc.).

Sub. A dummy variable taking a value of 1 if the winning firm subcon-tracts part of the work in a project, and 0 otherwise.

No. of subcontractors. The number of subcontractors working on aproject.

Value of subcontract. The value (in euros) of the subcontracts for aproject.

Bidder-subcontractor. A dummy variable that takes a value of 1 if, for agiven contract, at least one subcontractor participated as a bidder in theauction. It takes a value of 0 otherwise.

Type of CA. A set of dummy variables representing the type of contract-ing authority auctioning the work (e.g., regional or local governments,public health authorities, etc.).


ownloaded from



I user on 21 Decem

ber 2020

Appendix B

ReferencesAlbano, G., M. Bianchi, and G. Spagnolo. 2006. “Bid Average Methods in procurement,”

1–2 Rivista di Politica Economica 41–62.

Athey, S., D. Coey, and J. Levin. 2013. “Set-Asides and Subsidies in Auctions,” 5 American

Economic Journal: Microeconomics 1–27.

Bajari, P., S. Houghton, and S. Tadelis. 2014. “Bidding for Incomplete Contracts: an

Empirical Analysis of Adaptation Costs,” 104 American Economic Review 1288–319.

Bajari, P., R. McMillan, and S. Tadelis. 2009. “Auctions versus Negotiations in

Procurement: an Empirical Analysis,” 25 Journal of Law, Economics and Organization

372–99.

Bucciol, A., O. Chillemi, and G. Palazzi. 2013. “Cost Overrun and Auction Format in Small

Size Public Works,” 30 European Journal of Political Economy 35–42.

Chang, W.S., B. Chen, and T.C. Salmon. 2013. “An Investigation of the Average Bid

Mechanism for Procurement Auctions,”Management Science forthcoming.

Coviello, D., and S. Gagliarducci. 2010. “Building Political Collusion: Evidence from

Procurement Auctions,” IZA Discussion Papers 4939, Institute for the Study of Labor

(IZA).

Decarolis, F. 2009. “When the Highest Bidder Loses the Auction: Theory and Evidence from

Public Procurement,” Economic working papers 717, Bank of Italy, Economic Research

Department.

———. 2013. “Awarding Price, Contract Performance and Bids Screening: Evidence from

Procurement Auctions,” 6 American Economic Journal: Applied Economics 108–32.

De Silva, D.G., T. Dunne, G. Kosmopoulou, and C. Lamarche. 2012. “Disadvantaged

Business Enterprise Goals in Government Procurement Contracting: An Analysis of

Bidding Behavior and Costs,” 4 International Journal of Industrial Organization 377–88.

Galavotti, S., L. Moretti, and P. Valbonesi. 2014. “Sophisticated Bidders in Beauty Contest

Auctions,” Marco Fanno Working Paper 187, Department of Economics and

Management, University of Padova.

| | | |

winning bid

xamnim

10°p A1 A2 90°p

Area A

Average bid auc�on

Average bid auc�on + lo�ery

min max

10°p A3 A1 A2 90°p

2 4 6 81 3 5 7 9

winning bid

Area A

W if R=5

| | | | |

Figure B1. Types of auction.


ownloaded from



I user on 21 Decem

ber 2020

Gale, I.L., D.B. Hausch, and M. Stegeman. 2000. “Sequential Procurement with

Subcontracting,” 41 International Economic Review 989–1020.

Gil, R., and J. Marion. 2013. “Self-Enforcing Agreements and Relational Contracting:

Evidence from California Highway Procurement,” 29 Journal of Law, Economics, and

Organization 239–77.

Grossman, S., and O. Hart. 1986. “The Costs and Benefits of Ownership: a Theory of

Vertical and Lateral Integration,” 94 Journal of Political Economy 691–719.

Grossman, G.M., and E. Helpman. 2002. “Integration versus Outsourcing in Industry

Equilibrium,” 117 Quarterly Journal of Economics 85–120.

Joskow, P.L. 1988. “Asset Specificity and the Structure of Vertical Relationships: Empirical

Evidence,” 4 Journal of Law, Economics and Organization 95–117.

Kamien, M., and L. Li. 1990. “Subcontracting, Coordination, Flexibility and Production

Smoothing in Aggregate Planning,” 36 Management Science 1352–63.

Kamien, M., L. Li, and D. Samet. 1989. “Bertrand Competition with Subcontracting,” 20

Rand Journal of Economics 553–67.

Kogut, B., and U. Zander. 1992. “Knowledge of the Firm, Combinative Capabilities, and the

Replication of Technology,” 3 Organization Science 383–97.

———. 1996. “What Firms Do? Coordination, Identity, and Learning,” 7 Organization

Science 502–18.

Krasnokutskaya, E., and K. Seim. 2011. “Bid Preference Programs and Participation in

Highway Procurement Auctions,” 101 American Economic Review 2653–86.

Lafontaine, F., and M. Slade. 2007. “Vertical Integration and Firm Boundaries: The

Evidence,” 45 Journal of Economic Literature 629–85.

———. 2008. “Exclusive Contracts and Vertical Restraints: Empirical Evidence and Public

Policy,” in P. Buccirossi, ed.,Hanbook of Antistrust Economics. 391–414. CambridgeMA:

MIT Press.

Lewis, T.R., and D.E.M. Sappington. 1991. “Technological Change and the Boundaries of

the Firm,” 81 American Economic Review 887–900.

Lewis, G., and P. Bajari. 2011. “Procurement Contracting with Time Incentives: Theory and

Evidence,” 126 Quarterly Journal of Economics 1173–211.

Macher, J.T. 2006. “Technological Development and the Boundaries of the Firm: A

Knowledge-Based Examination in Semiconductor Manufacturing,” 52 Management

Science 826–43.

Marion, J. 2007. “Are Bid Preferences Benign? The effect of Small Business Subsidies in

Highway Procurement Auctions,” 91 Journal of Public Economics 1591–624.

———. 2009. “How Costly is Affirmative Auction? Government Contracting and

California’s Proposition 209,” 91 Review of Economics and Statistics 503–22.

———. 2013. “Sourcing from the Enemy: Horizontal Subcontracting in Highway

Procurement,”The Journal of Industrial Economics forthcoming.

Miller, D. 2014. “Subcontracting and Competitive Bidding on Incomplete Procurement

Contracts,” 45 The Rand Journal of Economics 705–46.

Novak, S., and S. Stern. 2008. “How Does Outsourcing Affect Performance Dynamics?

Evidence from the Automobile Industry,” 54 Management Science 1963–79.

OECD. 2007. Integrity in Public Procurement; available at: www.oecd.org/development/

aideffectiveness/38588964.pdf.

Olivares, M., G.Y. Weintraub, R. Epstein, and D. Yung. 2012. “Combinatorial Auctions for

Procurement: An Empirical Study of the Chilean SchoolMeals Auction,” 58Management

Science 1458–81.

Quinn, J.B., and F.G. Hilmer. 1994. “Strategic Outsourcing,” Summer Sloan Management

Review 43–55.

Riordan, M., and D. Sappington. 1987. “Information, Incentives, and Organizational

Mode,” 102 Quarterly Journal of Economics 243–264.

Spagnolo, G. 2012. “Reputation, Competition, and Entry in Procurement,” 30 International

Journal of Industrial Organization 291–6.

Spiegel, Y. 1993. “Horizontal Subcontracting,” 24 Rand Journal of Economics 570–90.


ownloaded from



I user on 21 Decem

ber 2020

www.oecd.org/development/aideffectiveness/38588964.pdf

www.oecd.org/development/aideffectiveness/38588964.pdf

Van Mieghem, J.A. 1999. “Coordinating Investment, Production and Subcontracting,” 45

Management Science 954–71.

Wan, Z., and D.R. Beil. 2009. “RFQ Auctions with Supplier Qualification Screening,” 57

Operational Research 934–49.

Wan, Z., D.R. Beil, and E. Katok. 2012. “When Does It Pay to Delay Supplier Qualification?

Theory and Experiment,” 58 Management Science 2057–75.

Webster, M., C. Alder, and A.P. Muhlemann. 1997. “Subcontracting within the Supply

Chain for Electronics Assembly Manufacture,” 17 International Journal of Operations

and Production Management 827–84.

Williamson, O. 1971. “The Vertical Integration of Production: Market Failure

Considerations,” 63 American Economic Review 316–25.

———. 1985. The Economic Institutions of Capitalism. New York: The New Press.


ownloaded from



I user on 21 Decem

ber 2020

jleo, v31 n3 firms’ qualifications and subcontracting in

Documents